PovelikinRostislav

from __future__ import print_function
import argparse
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from torch.optim.lr_scheduler import StepLR

PovelikinRostislav

Back Propagation

Устройство сети

ИЛЛЮСТРАЦИЯ СЕТЕЙ - по ссылкам в комментарии.

• 2 нейрона входа
• 3 нейрона выхода, SoftMax
• Cross Entropy loss

W =

PovelikinRostislav

#include <tuple>

template<typename Functor> struct functor_arg : functor_arg<decltype(&Functor::operator())> {};

template<typename Functor, typename ReturnType, typename... ArgTypes> struct functor_arg<ReturnType (Functor::*)(ArgTypes...) const> {
    template<std::size_t N> using arg_t = std::tuple_element_t<N, std::tuple<ArgTypes...>>;
};

int main() {
    auto func = [](int) {};